Lymphedema pathological network construction method based on graph neural network
Through image processing and multi-layer graph convolution network model, combined with multimodal data fusion and time series analysis, the problems of data scarcity and computational complexity in the construction of lymphedema pathological network are solved, and efficient pathological change identification and disease are achieved. Tracking enhances the robustness and explanatory nature of the model.
Patent Information
- Application Number
- CN202410941261.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-15
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-07-15
AI Technical Summary
The existing lymphedema pathological network construction method relies on complex graph neural networks, requires a large amount of labeled data, and has high computational complexity, making it difficult to apply in resource-limited environments, and there are data scarcity, bias and insufficient model interpretation. question.
By collecting medical image data and clinical records, using image processing technology to extract key features, define the topology of the graph, designing multi-layer graph convolution network models, introducing attention mechanisms, combining multimodal data fusion and time series analysis, development based on The abnormality detection algorithm of the graph is used to identify pathological changes in lymphedema.
It improves the early recognition ability of lymphedema, is suitable for processing non-Euclidean structural data, can track the development of the disease, enhances the understanding and classification accuracy of pathological status, and reduces the dependence on professional knowledge.
Smart Images

Figure CN118919099A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for constructing a lymphedema pathology network, and in particular to a method for constructing a lymphedema pathology network based on a graph neural network. Background Art
[0002] Although current methods for constructing lymphedema pathology networks offer a variety of potential advantages, such as enhanced data analysis capabilities and improved pathology recognition accuracy, they also have a series of shortcomings and drawbacks. First, the construction of lymphedema pathology networks relies on highly complex graph neural network (GNN) models, which require a large amount of labeled data for training. In the actual medical field, especially when dealing with specific types of lymphedema such as breast cancer-related lymphedema, it is difficult to obtain sufficient clinical data to train these models. The scarcity of data not only limits the training quality of the model, but also leads to overfitting, that is, the model performs well on the training data but poorly on new unseen data. In addition, these data often involve privacy issues and must strictly comply with medical data protection regulations, which brings additional challenges in data collection and processing. Second, the computational complexity of graph neural networks is high, especially when processing large-scale graph data, which requires a lot of computing resources. This not only increases the cost of research, but also limits the application of these models in resource-limited settings. In low-resource medical environments, there is a lack of sufficient technical and computing support to run complex graph analysis models. In addition, the training and inference time of GNN models is very long, which is not practical in clinical environments that require rapid diagnosis and treatment decisions.
[0003] Furthermore, graph neural networks are deficient in model interpretability. Although they perform well in feature extraction and pattern recognition, it is difficult to explain the specific behavior and decision path of the model due to the complexity of the model structure. In the medical field, the interpretability of the model is very important, and doctors and patients usually need to understand the reasoning process of the model to increase trust and acceptance. Lack of transparency will hinder the acceptance and application of graph-based methods in clinical practice. In addition, the current lymphedema pathology network construction method is biased in patient groups of different races, ages or genders, which is caused by the unevenness in the training data set. If the data used for training does not represent a wide range of people, the model will perform well in specific subsets and poorly in other subsets. This bias can be alleviated by using more diverse and inclusive datasets, but such dataset construction usually requires additional time and resources. In addition, although graph neural networks have significant advantages in recognizing complex patterns, they are very sensitive to noise and outliers in the data. In actual clinical data, data quality problems caused by various factors (such as different equipment, operator technical level differences, etc.) are often encountered. These problems will affect the performance of graph neural networks, leading to misdiagnosis and missed diagnosis. Therefore, further research is needed on how to optimize models to adapt to fluctuations in data quality and develop new robust algorithms to handle imperfect input data. Summary of the invention
[0004] The purpose of the present invention is to provide a method for constructing a lymphedema pathology network based on a graph neural network, thereby solving some of the drawbacks and shortcomings pointed out in the background technology.
[0005] The present invention solves the above-mentioned technical problems by adopting the following technical solutions:
[0006] First, we collected the patient's lymphatic system medical imaging data, clinical records, and biomarker data, and used image processing technology to extract key features and structure them into nodes and edges of a graph, where nodes represent lymph nodes and edges represent the paths or interactions of lymphatic flow.
[0007] Then, based on the anatomical structure and physiological function of the lymphatic system, the topological structure of the graph is defined, and multimodal data fusion technology is introduced to combine imaging data and biomarker data to enhance the information accuracy of the graph network;
[0008] Then, a multi-layer graph convolutional network model is designed to learn the patterns of lymph nodes and their connections, and an attention mechanism is introduced to identify and emphasize nodes and edges that play a role in the development of lymphedema.
[0009] Then, deep learning is performed using graph convolutional networks to extract hidden features of nodes and edges, and graph embedding techniques are applied to transform high-dimensional graph data into low-dimensional space for classification or clustering analysis.
[0010] Finally, a graph-based anomaly detection algorithm was developed to identify pathological changes that were significantly different from common patterns and combined with time series analysis to predict the pathological progression of lymphedema.
[0011] Furthermore, the extracting key features using image processing technology comprises the following steps:
[0012] S1. First, adopt a unified data processing framework:
[0013]
[0014] G and H represent processing functions for different data sources including CT and MRI imaging data, blood biochemical indicators, and patient symptom records, which are used to synchronously process and integrate multiple types of data;
[0015] S2, then apply the image processing algorithm through the function:
[0016]
[0017] where v n is the value of each pixel in the image, is the average pixel value, which is used to identify and quantify the size, shape, and distribution characteristics of lymph nodes;
[0018] S3, then build the algorithm through the graph:
[0019]
[0020] Where E is the set of edges, α ij is the edge weight, k i and k j It is the characteristic value of the nodes connecting the edges, which is used to convert the extracted features into nodes in the graph, i.e., representing lymph nodes, and edges, i.e., representing the physiological relationship or signal transmission path between nodes.
[0021] Furthermore, the topological structure of the definition graph includes the following steps:
[0022] S1. First, use the graph topology generation algorithm:
[0023]
[0024] Where N represents nodes, L represents edges, and L i is a connection specified based on anatomical structure, A i is the corresponding physiological functional connection, d is the distance between nodes, and σ is the normalization parameter used to construct the nodes and edges of the graph according to the anatomical location and physiological function of the lymph nodes;
[0025] S2, then apply the multimodal deep fusion network through the function:
[0026]
[0027] Where x, y, z represent the input data of different modes, w k and v j It is the weight learned from the data, which is used to integrate information from different sources including MRI, CT scans, and biochemical markers to enhance the accuracy of the final graph structure data.
[0028] Furthermore, the multi-layer graph convolutional network model construction process:
[0029] First, we introduce and implement the adaptive layer structure of the multi-layer graph convolutional network GCN, using the calculation formula:
[0030]
[0031] Where G represents the overall structure of the graph, E represents the edge set in the graph, and x i and x i-1 Represent the feature vectors of consecutive nodes, α i is the inter-layer adaptability factor, ∈ is the smoothing constant, which is used to adjust the structure of the network layer to adapt to the anatomical characteristics and pathological changes of the lymph nodes;
[0032] Secondly, the graph-based attention mechanism is introduced, and the calculation formula is:
[0033]
[0034] where u and v are nodes in the graph, f(u) and f(v) represent the feature vectors of nodes u and v, β is the learned attention concentration parameter, and δ is a regularization constant used to emphasize nodes and edges that are critically related to the development of lymphedema.
[0035] Furthermore, the method for extracting hidden features of nodes and edges adopts the steps of:
[0036] S1. First, we design and implement an adaptive layer structure of a multi-layer graph convolutional network (GCN). We adapt the connectivity and feature diversity of nodes by adjusting the graph convolution operation of the convolution kernel parameters. The convolution kernel size and feature extraction method of each layer are dynamically adjusted according to the local structure of the graph to capture the biomarkers and pathological characteristics between lymph nodes.
[0037] S2. Further integrate multi-scale feature extraction strategies to capture biomarker features from micro to macro at different levels, and introduce cross-layer connections to enhance information flow and feature reuse to improve the accuracy of classification and clustering;
[0038] S3. Finally, a nonlinear dimensionality reduction method based on deep learning is applied for graph embedding. Designed for graph data, while maintaining the topological structure of the graph and the relationship between nodes, a high-dimensional to low-dimensional mapping is constructed to facilitate subsequent classification or clustering analysis.
[0039] Furthermore, the adaptive layer structure adjusts the function through a dynamic convolution kernel:
[0040]
[0041] Implementation, where G represents the overall structure of the graph, E represents the edge set, and α vu represents the adaptive weight on the edge (v,u), x v and x u denote the feature vectors of nodes v and u, respectively. θ is a parameter adjusted based on the topology of the graph and the diversity of node features to adapt to the feature extraction requirements of each layer, especially in terms of the biomarkers of lymph nodes and their interactions.
[0042] Furthermore, the multi-scale feature integration technology is used to effectively capture the specific biomarkers and pathological features of lymphedema, using the formula:
[0043]
[0044] Among them, H l represents the output feature matrix of the lth layer, W k represents the weight matrix from layer k to layer l, and σ is the nonlinear activation function used to integrate information at multiple levels of the graph and extract biomarkers.
[0045] Furthermore, the integrated multi-scale feature extraction strategy includes the following steps:
[0046] S1. First, implement the multi-scale graph convolutional network M-GCN through the function:
[0047]
[0048] Where v represents the current node, N k (v) is the set of neighbor nodes of node v at the kth scale, ω k is the learning weight of the kth scale, x u is the feature of the neighbor nodes, σ is the activation function designed to capture the biomarker features from micro to macro at different levels;
[0049] S2, then introduce cross-layer connections to enhance information flow and feature reuse, specifically through the formula:
[0050]
[0051] Implementation, where F l is the output feature of the lth layer, H i is the output feature of the i-th layer, γ li It is the cross-layer connection weight from the i-th layer to the l-th layer, which is used to improve the accuracy of classification and clustering.
[0052] Furthermore, the nonlinear dimensionality reduction method for constructing a graph embedding framework includes:
[0053] S1. First, we implement the deep learning driven graph embedding model DLGE, which learns the nonlinear embedding of the graph through the autoencoder, specifically by minimizing the function:
[0054]
[0055] Where V and E represent the node set and edge set in the graph respectively, f(v) represents the low-dimensional embedding of node v, and λ is the regularization coefficient, which is used to adjust the penalty strength of non-adjacent node pairs to maintain the topological structure of the graph and the relationship between nodes in the process of high-dimensional to low-dimensional mapping;
[0056] S2. Then introduce the graph structure preservation loss function GSPL, by calculating:
[0057]
[0058] where ω ij is the importance weight of the node pair (i, j) learned based on the graph attention layer, x i and x j is the original feature vector, g(i) and g(j) are the corresponding low-dimensional representations.
[0059] Furthermore, the graph-based anomaly detection algorithm consists of:
[0060] S1. First, implement the dynamic graph convolution network DGCN, using the dynamic graph convolution layer, through the function:
[0061]
[0062] To update the node representation, where t represents the time step, v is the current node, is the neighbor set of node v, W t and b t are the weight matrix and bias vector at time step t, respectively, h t-1 (u) is the embedding representation of node u at time t-1, σ is the standard deviation of the Gaussian kernel, which is used to model the interaction strength between nodes;
[0063] S2, then apply the graph-based anomaly detection algorithm GAD, by calculating:
[0064]
[0065] to identify anomalies, where f(v) and f(u) are the embedding vectors of node v and its neighbor u, respectively, and p is a positive real number selected to increase the sensitivity and discrimination of anomaly detection;
[0066] S3. Finally, combining time series analysis and prediction model, the time series graph data is processed through the long short-term memory network LSTM. The specific implementation formula is:
[0067] h t =LSTM(h t-1 ,α·h t-1 +(1-α)·x t )
[0068] where h t is the hidden state at time t, x t is the input feature, and α is a learning rate between 0 and 1, which is used to balance the influence of historical information and current input to predict the pathological progression of lymphedema.
[0069] Beneficial effects of the present invention:
[0070] 1. Use graph neural networks (GNNs) to perform deep analysis of pathological data. By capturing the interactions and patterns between lymph nodes, these networks can more accurately identify various biomarkers and pathological changes of lymphedema. GNN's highly complex data processing capabilities improve the ability to identify early signs of the disease, which is crucial for early diagnosis and timely treatment.
[0071] 2. Graph-based methods are particularly suitable for processing non-Euclidean structured data, including medical images and pathological networks. GNNs can learn and extract key features from data without relying on traditional, handcrafted feature extraction techniques. This not only reduces the reliance on expertise, but also improves the efficiency and effectiveness of processing complex data.
[0072] 3. Through the dynamic graph convolutional network, the present invention is able to process pathological features that change over time, making it possible to track the development and changes of the disease. This is particularly useful for disease states that require continuous monitoring and regular evaluation, such as the dynamic changes of lymphedema during treatment.
[0073] 4. By implementing a multi-scale feature integration strategy, biomarker features from micro to macro can be captured at different levels. Such multi-level analysis enhances the understanding of the complexity of pathological states, thereby improving the accuracy of classification and clustering. BRIEF DESCRIPTION OF THE DRAWINGS
[0074] Figure 1This is a flow chart of the lymphedema pathology network construction method based on graph neural network of the present invention.
[0075] Figure 2 This is a flow chart of the method for extracting hidden features of nodes and edges according to the present invention. DETAILED DESCRIPTION
[0076] The specific implementation modes of the present invention will be described in detail below in conjunction with the accompanying drawings.
[0077] Attached to this case Figure 1 In the first step, data including medical images (such as CT and MRI scans), clinical records and biomarkers (such as blood test results) are collected from patients. Through advanced image processing techniques, edge detection and texture analysis, key features are extracted from medical images, including the size, shape and distribution of lymph nodes, and these data points are converted into nodes of the graph.
[0078] Attached to this case Figure 1 In the second step, the edges connecting these nodes are then defined according to the lymphatic flow paths and the physiological interactions between the lymph nodes. Furthermore, the topology of the graph is defined using the anatomical and physiological structure knowledge of the lymphatic system, and multimodal data fusion technology is applied to enhance the overall information accuracy and richness of the graph network by integrating information from different data sources.
[0079] Attached to this case Figure 1 In the third step of the paper, a multi-layer graph convolutional network (GCN) model is then designed and implemented to learn and identify the patterns of lymph nodes and their interconnections by processing the graph data layer by layer. In this process, an attention mechanism is introduced to identify and emphasize the nodes and edges that play a key role in the development of lymphedema, thereby more accurately capturing the development trend of the pathological state.
[0080] Attached to this case Figure 1 In the fourth step, graph convolutional network deep learning is used to extract hidden features of nodes and edges, and these complex high-dimensional graph data are effectively converted into low-dimensional representations through graph embedding technology for more effective data analysis, including classification and clustering.
[0081] Attached to this case Figure 1 The 5 steps in the study are to develop a graph-based anomaly detection algorithm specifically designed to identify changes that are significantly different from known and common lymph node pathology patterns, thereby providing early warning of potential pathological abnormalities. This algorithm, combined with time series analysis techniques, can not only detect the current abnormal state, but also predict the future development trend of lymphedema.
[0082] Embodiment 1:
[0083] The embodiment first needs to collect data from patients, including CT scans, MRI imaging data, blood biochemical indicators, and clinical symptom records of patients. These data exist in different data formats and scales, so a unified data processing framework is needed to integrate this information for further analysis. The following mathematical model is used to integrate the data:
[0084]
[0085] Among them, G(x i ,y i ,z i ) and H(x i ,y i ,z i ) represent processing functions for various medical images and biochemical indicators. G function represents feature extraction algorithm, including algorithms for extracting lymph node size and shape from image data, while H function is a normalization function used to standardize data to eliminate dimension effects and differences between different devices.
[0086] The G and H functions are defined as:
[0087]
[0088] H(x i ,y i ,z i )=x i +y i +z i
[0089] where x i ,y i ,z i are the normalized values of the 3D coordinate data extracted from the CT scan. These values vary between 0 and 1 and represent the relative position of the lymph nodes in the body.
[0090] Assume that the 3D coordinate data extracted from the CT scan has been normalized to the range of 0,1, and the specific value is x i =0.5,y i =0.3,z i = 0.4. Using these values into the above function, first calculate the values of G and H:
[0091] G(0.5,0.3,0.4)=0.5 2 +0.3 2 +0.4 2 =0.25+0.09+0.16=0.50
[0092] H(0.5,0.3,0.4)=0.5+0.3+0.4=1.2
[0093] Further, the integral calculation is performed:
[0094]
[0095] This value can be viewed as an integrated measure of the single point estimate of the given node feature in the entire graph, which will be used to set the weights of the graph nodes and as a feature input into the graph neural network.
[0096] Further technical implementations include applying this type of processing to all collected data points, creating a complete graph representation, where each node represents a specific lymph node, and the attributes of the node are derived from the results of similar processing as described above. The definition of edges is based on the paths or interactions of lymphatic flow, which can be obtained from clinical records and input through expert systems. Once the graph is created, a multi-layer graph convolutional network model will be trained to recognize and learn the patterns and connections between lymph nodes.
[0097] The embodiment then uses an image processing algorithm to identify and quantify the size, shape and distribution characteristics of lymph nodes, involving using the image processing algorithm to perform feature extraction through the following function:
[0098]
[0099] where v n is the value of each pixel in the image, and is the average of these pixel values. This function calculates the root mean square difference between the image pixel value and the average pixel value, which can be used to quantify the visual differences of objects in the image, including the size and morphological characteristics of lymph nodes.
[0100] To further specify and verify the feasibility of this scheme, assume that an image region containing lymph nodes is obtained from a CT scan, and the region contains N = 100 pixels. The values vn of these pixels in the grayscale image range from 0 to 255, and these pixel values are assumed to be randomly distributed, with the following specific values (example values):
[0101] v = [30, 45, 60, 120, 150, 170, 180, 190, 200, 210] (note that in practice there will be 100 such values, only 10 are shown here to simplify the calculation).
[0102] Average pixel value The calculation is as follows:
[0103]
[0104] Then substitute each pixel value into the above formula and calculate:
[0105]
[0106] M(v)=60
[0107] This result provides a quantitative measure indicating that there is a clear difference in pixel values between the lymph nodes and the background in the image, with higher values indicating a higher contrast between the lymph nodes and the surrounding tissue, which is an indicator of abnormal lymph node size and shape.
[0108] The final embodiment uses the extracted features to construct a detailed graph representation for the analysis of lymphedema pathological network, which involves converting the extracted image features into nodes (representing lymph nodes) and edges (representing physiological relationships or signal transmission paths between nodes) in the graph, which is achieved through the following construction algorithm:
[0109]
[0110] Where E is the set of edges, α ij is the edge weight, k i and k j is the eigenvalue of the nodes connecting the edges.
[0111] In the medical image, several key lymph nodes have been identified through the above image processing steps, and their eigenvalues (such as metrics based on size, shape, and contrast extraction) are k1=150, k2=180, and k3=160 respectively. These eigenvalues represent quantitative indicators of certain physiological characteristics and pathological states of lymph nodes. Furthermore, the connection relationship between these nodes is set to be based on the physical distance in the body and the similarity of physiological functions, and the corresponding edge weight α ij It can be set to be determined based on the importance of these physiological and pathological relationships, with weights ranging from 0.1 to 1, depending on the physiological significance of the connection and the influence of the pathological state.
[0112] Based on this, the connection between lymph node 1 and lymph node 2 is very important (showing a high correlation in the pathological process), and α is set 12 =1; for relatively less important connections (such as lymph nodes 1 and 3), set α 13 = 0.5. Substituting these values into the above formula, we can calculate the structural loss value for this graph:
[0113] L(E)=α 12 (k1-k2) 2 +α 13 (k1-k3) 2
[0114] L(E)=1×(150-180) 2 +0.5×(150-160) 2
[0115] L(E)=1×900+0.5×100
[0116] L(E)=900+50
[0117] L(E)=950
[0118] The computational results provide quantitative indicators that reflect the overall strength and stability of the relationships between nodes (lymph nodes) in the graph based on given features, helping researchers understand which connections between lymph nodes are particularly important and abnormal in the pathological process, thereby guiding the diagnosis and treatment strategies of lymphedema.
[0119] Embodiment 2:
[0120] In this embodiment, the topological structure of the definition graph is constructed, which involves a graph topology generation algorithm, which is expressed by the following formula:
[0121]
[0122] In order to illustrate the implementation steps and feasibility of this solution in detail, it is assumed that several key lymph nodes have been identified from medical images when constructing the lymphedema pathology network. These nodes need to be connected to each other according to their anatomical location and physiological function. Consider the following specific data and parameter settings:
[0123] N represents the node set, and there are three lymph nodes, namely node 1, node 2 and node 3;
[0124] L represents the edge set. According to the anatomical location and physiological function of lymph nodes, it is assumed that there is a connection between node 1 and node 2, and between node 2 and node 3;
[0125] d is the distance between nodes, which is set to 2 cm between nodes 1 and 2 and 3, based on the anatomical image analysis;
[0126] A i is the corresponding physiological functional connection. According to the physiological function analysis, the functional connection strength score between node 1 and node 2 is 0.8, and the functional connection strength score between node 2 and node 3 is 0.6;
[0127] Set the value of σ to 1 to balance the scale in the distance calculation.
[0128] Based on these settings, the weight of each pair of connections can be calculated:
[0129]
[0130] F(N,L)=exp(-0.72)+exp(-2.88)
[0131] F(N,L)=0.486+0.056
[0132] F(N,L)=0.542
[0133] The computational results provide quantitative metrics showing the overall weight of the edges in the graph constructed based on anatomical and physiological data, ensuring that the graph model not only reflects the physical proximity of lymph nodes but also takes into account interactions in physiological functions.
[0134] In the next step, a multimodal deep fusion network is used to integrate information from different data sources to enhance the accuracy of graph structure data. By extracting and merging information from multiple different data modalities such as MRI, CT scans, and biochemical markers, a more comprehensive and richer feature representation is provided for the graph neural network. The following multimodal fusion function is used for data integration:
[0135]
[0136] Where 1 represents the input data from different modalities, w k and v j are the weights learned from the data.
[0137] In the actual application, data is extracted from MRI scans (x data modality), CT scans (y data modality), and blood biochemical markers (z data modality) of patients.
[0138] x=[x1,x2,…,x K ] represents the feature vector extracted from the MRI scan, including features related to lymph node size;
[0139] y=[y1,y2,…,y J ] represents the feature vector extracted from the CT scan, including features related to lymph node density;
[0140] z represents continuous data extracted from blood biochemical markers, including the concentration of specific proteins in the blood.
[0141] There are 5 MRI features and 3 CT features, with weights w k and v j Learned by data-driven method. The specific weight value can be automatically adjusted according to the importance during training, ranging from 0.1 to 1. For simplicity, all weights are set to 0.5.
[0142] Substitute the data and weights into the fusion function and perform the following calculations:
[0143] G(x,y,z)=(0.5×x1+0.5×x2+…+0.5×x5)·(0.5×y1+0.5×y2+0.5×y3)+∫z dz
[0144] Set x i =100 and y j =200 is a simplified calculation value, and the integral of z is 300:
[0145] G(x,y,z)=(0.5×500)·(0.5×600)+300
[0146] G(x,y,z)=250×300+300
[0147] G(x,y,z)=75300
[0148] The calculation results integrate the image features of MRI and CT with biochemical marker data, making the final graph structure data more accurate and providing effective input for lymphedema pathology network analysis based on graph neural network.
[0149] Embodiment 3:
[0150] The embodiment constructs an adaptive layer structure of a multi-layer graph convolutional network (GCN) for capturing and analyzing the anatomical features and pathological changes of lymph nodes. The process adjusts the network structure according to the feature differences between nodes through a calculation formula, so that the model can better adapt to complex medical data and pathological conditions, as follows:
[0151]
[0152] Where G represents the overall structure of the graph, E represents the edge set in the graph, and x i and x i-1 Represent the feature vectors of consecutive nodes, α i is the inter-layer adaptability factor, and ε is a smoothing constant used to adjust the structure of the network layer.
[0153] Consider a small network with four nodes, each node represents a lymph node, and the node's feature vector includes parameters such as the size, shape, and density of the lymph node. The specific values are as follows:
[0154] x1=[1.0,0.8]
[0155] x2=[1.2,0.9]
[0156] x3=[1.1,0.85]
[0157] x4=[1.3,0.95]
[0158] Setting α iThe value range is 0.1 to 1.0, and the specific value can be automatically adjusted according to the learning importance of each layer. In this example, all α i The values are all 0.5, and ∈ is set to 0.01 to reduce the impact caused by small feature differences.
[0159] The term corresponding to the feature difference between each node is calculated according to the formula:
[0160]
[0161] Substituting these values into the overall formula:
[0162] F(G,E)=0.5×log(6)+0.5×log(3.75)+0.5×log(5.5)
[0163] F(G,E)≈0.5×1.792+0.5×1.321+0.5×1.704
[0164] F(G,E)≈0.896+0.660+0.852
[0165] F(G,E)≈2.408
[0166] The computational results provide quantitative measures of how the model adjusts its internal structure based on the differences between feature vectors, optimizing information flow and feature learning.
[0167] The next step in the implementation example is to introduce a graph-based attention mechanism, with the goal of identifying and emphasizing nodes and edges that play a key role in the development of lymphedema, thereby providing in-depth insights into the dynamic changes of the disease, which is achieved through the following calculation formula:
[0168]
[0169] where u and v are nodes in the graph, f(u) and f(v) represent the feature vectors of nodes u and v respectively, β is the learned attention concentration parameter, and δ is the regularization constant.
[0170] Nodes u and v are set to represent two lymph nodes, and their feature vectors are f(u) = [1.2, 0.8] and f(v) = [1.1, 0.9], respectively. β is set to 2.0, reflecting the strong focus on the differences between node features in the analysis. Similarly, the regularization constant δ is set to 0.5 to adjust the impact scale of feature differences.
[0171] Substitute the formula for calculation and first solve the square of the difference between the two node eigenvectors:
[0172] ||f(u)-f(v)|| 2 =(1.2-1.1) 2+(0.8-0.9) 2 =0.01+0.01=0.02
[0173] Then substitute the attention calculation formula:
[0174]
[0175] A(u,v)=exp(-0.08)≈0.923
[0176] This calculation result shows that there is a high attention weight between nodes u and v, which means that the interaction between the two lymph nodes in the pathological network is relatively important and plays a key role in the pathological development of lymphedema.
[0177] Embodiment 4:
[0178] In developing a method for constructing a lymphedema pathology network based on a graph neural network, one of the core links is to design and implement an adaptive layer structure of a multi-layer graph convolutional network (GCN). The dynamic convolution kernel adjustment function is used to adapt to the connectivity and feature diversity between nodes, ensuring that the convolution kernel size and feature extraction method of each layer can be dynamically adjusted according to the local structure of the graph, thereby effectively capturing the biomarkers and pathological features between lymph nodes.
[0179] The key technologies involved in implementing this solution are:
[0180]
[0181] In this formula, G represents the overall structure of the graph, E represents the edge set, and α vu represents the adaptive weight on the edge (v,u), x v and x u denote the feature vectors of nodes v and u respectively, and θ is an adjustment parameter used to adapt the feature extraction requirements of each layer according to the topology of the graph and the diversity of node features.
[0182] In order to demonstrate the implementation process and feasibility of this scheme, a specific example is considered. A simplified lymphedema pathology network is set, including three nodes and two edges. The feature vector of the node is expressed as:
[0183] x1=[1.0,2.0]
[0184] x2=[1.5,1.8]
[0185] x3=[0.9,2.1]
[0186] The set of edges and the corresponding adaptive weights (set between 0.1 and 1.0, with specific values determined according to the physiological importance of the edge) are:
[0187] E={(1,2),(2,3)}
[0188] α 12 =0.9,α 23 =0.7
[0189] Set θ to 0.5 to emphasize the feature differences between connections, and apply the above function to calculate the adjusted network structure:
[0190] F(G,θ)=0.9exp(-0.5||[1.0,2.0]-[1.5,1.8]|| 2 )+0.7exp(-0.5||[1.5,1.8]-[0.9,2.1]|| 2 )
[0191] =0.9exp(-0.5((1.0-1.5) 2 +(2.0-1.8) 2 ))+0.7exp(-0.5((1.5-0.9) 2 +(1.8-2.1) 2 ))
[0192] =0.9exp(-0.5(0.25+0.04))+0.7exp(-0.5(0.36+0.09))
[0193] =0.9exp(-0.5×0.29)+0.7exp(-0.5×0.45)
[0194] =0.9exp(-0.145)+0.7exp(-0.225)
[0195] =0.9×0.865+0.7×0.799
[0196] =0.7785+0.5593
[0197] =1.3378
[0198] Computational results show that the dynamically adjusted graph convolutional network can effectively capture and emphasize important biomarkers and pathological changes.
[0199] The embodiment further applies multi-scale feature integration technology to enable the network to capture and fuse biomarkers and pathological features at different levels, thereby more comprehensively understanding the complex interactions between lymph nodes and their changes during pathological processes. The following formula is used to integrate features and promote deep learning of information:
[0200]
[0201] H lrepresents the output feature matrix of the lth layer, W k is the weight matrix from the kth layer to the lth layer, and σ is a nonlinear activation function used to strengthen the nonlinear combination and extraction between features.
[0202] The graph neural network is assumed to have three layers, and the purpose of each layer is to capture and integrate information from different data sources (such as MRI, CT, biochemical data). The initial representation of the feature vector is directly extracted from the original medical images and biochemical data:
[0203] H 0 The input layer contains primary features, the specific value is H 0 =[0.2,0.8,0.5,0.9].
[0204] Next, set the weight matrix W k , which is different in each layer to adapt to the learning requirements of a specific layer. Set the weight matrix W of each layer k Initialized randomly and gradually adjusted and optimized through the training process, the initial weights are set as follows:
[0205] W 0 is the identity matrix I;
[0206] W 1 and 1 are the weights learned during the training process, and the initial setting is: W 1 =[0.5,0.4,0.6,0.7],W 2 =[0.3,0.7,0.5,0.2].
[0207] Use ReLU as the nonlinear activation function σ (ReLU function is a commonly used nonlinear activation function that effectively solves the gradient vanishing problem and provides model stability and fast convergence). Apply the formula to each layer and get:
[0208] H 1 =ReLU([1,0.5,0.6,0.7]×[0.2,0.8,0.5,0.9]+[0.5,0.4,0.6,0.7]×[0.2,0.8,0.5,0.9])
[0209] H 1 =ReLU(W 1 H 0 )=ReLU([0.5×0.2+0.4×0.8+0.6×0.5+0.7×0.9])
[0210] H 1 =ReLU([0.1+0.32+0.3+0.63])=ReLU([1.35])=[1.35] (Here the vector operation is set to accumulation)
[0211] Calculate H 2 :
[0212] H 2 =ReLU(W 0 H 1 +W 1 H 2 +W 2 H 0 )
[0213] H 2 =ReLU([1×1.35+0.5×1.35+0.3×0.2+0.7×0.8+0.5×0.5+0.2×0.9])
[0214] H 2 =ReLU([1.35+0.675+0.06+0.56+0.25+0.18])
[0215] H 2 =ReLU([3.075]) = [3.075]
[0216] This approach allows the system to capture the characteristics of pathological changes from local to global in each layer through hierarchical and nonlinear integration, thereby improving the accuracy of diagnosis and the depth of disease understanding.
[0217] Embodiment 5:
[0218] The embodiment implements an integrated multi-scale feature extraction strategy, aiming to capture biomarker features from micro to macro at different levels through a multi-layer graph convolutional network (M-GCN). This approach not only enhances the model's feature capture capability, but also enhances information flow and feature reuse through cross-layer connections, thereby improving the accuracy of classification and clustering. Use the following function to operate:
[0219]
[0220] Where v represents the current node, N k (v) is the set of neighbor nodes of node v at the kth scale, ω k is the learning weight of the kth scale, x u is the feature of the neighbor nodes, and σ is the activation function designed to capture the biomarker features from micro to macro at different levels.
[0221] The setting is processing a large graph containing hundreds of lymph nodes, each represented by a set of features defined by biomedical images and biochemical marker data, where each node represents a lymph node and each lymph node is connected to other lymph nodes that are anatomically adjacent to it.
[0222] Parameter and data settings:
[0223] Assume that each lymph node v has a set of features x v =[x v1 ,x v2 ,…,x vm ];
[0224] At different scales k, the neighbor set N k (v) including different numbers of neighbors, representing different spatial resolutions or functional groups;
[0225] Learning weight ω k From small to large range, such as 0.1 to 1.0, dynamically adjusted according to the importance of each scale.
[0226] Implement the calculation process:
[0227] Scale 1: Focus on the most direct physical neighbors, the 5 lymph nodes closest to v;
[0228] Scale 2: expanded to functionally relevant neighbors, 10 lymph nodes that share similar biochemical markers with v;
[0229] Scale 3: further expanded to 15 lymph nodes that exhibited similar dynamic changes during the pathological process.
[0230] At each scale, a new feature vector is computed for each node using the following steps:
[0231]
[0232] Among them, ω1, ω2, ω3 are weights of different scales, and σ uses the ReLU function for nonlinear transformation to enhance the expressive power of the model.
[0233] In this way, each node not only considers the influence of its most immediate neighbors, but also integrates biomarker and pathology information in the broader network, thereby providing more comprehensive biological and medical insights at different levels.
[0234] Next, the embodiment introduces cross-layer connections to significantly enhance information flow and feature reuse, thereby improving the classification and clustering accuracy of the entire network model. The implementation of this strategy is concretized by the following formula:
[0235]
[0236] F l is the output feature of the lth layer, H i is the output feature of the i-th layer, γ li is the cross-layer connection weight from layer i to layer l.
[0237] It is assumed that there are four layers in the constructed multi-layer graph convolutional network, i.e., l = 0, 1, 2, 3, and each layer produces a set of output features H 0 ,H 1 ,H 2 ,H 3 In order to achieve cross-layer connection, we first define the weight γ li , these weights determine the strength of the connection from early layers to later layers, so that the information of early layers can be reused in subsequent layers. For practical applications, the weights γ li The weights can range from 0.1 to 1.0, depending on the relative distances between layers or on parameters learned during network training.
[0238] Weight setting: Set γ li As follows, the weight calculation is simplified and the decreasing weight value is taken, indicating that the earlier layers contribute less to the later layers:
[0239] γ 10 =0.9,γ 20 =0.8,γ 21 =0.7,γ 30 =0.6,γ 31 =0.5,γ 32 =0.4
[0240] Layer output features: Set the output feature matrix of each layer as follows:
[0241] H 0 =[1.0,2.0]
[0242] H 1 =[1.5,2.5]
[0243] H 2 =[2.0,3.0]
[0244] H 3 =[2.5,3.5]
[0245] For F 1 :
[0246] F 1 =γ 10 ·H 0 =0.9×[1.0,2.0]=[0.9,1.8]
[0247] For F 2 :
[0248] F 2 =γ 20 ·H 0 +γ 21 ·H 1=0.8×[1.0,2.0]+0.7×[1.5,2.5]=[0.8+1.05,1.6+1.75]
[0249] =[1.85,3.35]
[0250] For F 3 :
[0251] F 3 =γ 30 ·H 0 +γ 31 ·H 1 +γ 32 ·H 2 =0.6×[1.0,2.0]+0.5×[1.5,2.5]+0.4×[2.0,3.0]
[0252] =[0.6+0.75+0.8,1.2+1.25+1.2]=[2.15,3.65]
[0253] Through such calculations, we can see that each layer effectively integrates the information from the previous layer, so that the features are not only captured within the current layer, but also enhanced by cross-layer connections.
[0254] Embodiment 6:
[0255] The embodiment applies a nonlinear dimensionality reduction method based on deep learning for graph embedding, converting complex high-dimensional graph data into a lower-dimensional space while maintaining the topological structure of the original graph and the relationship between nodes, thereby facilitating more accurate classification or clustering analysis. The nonlinear dimensionality reduction method is implemented by a deep learning driven graph embedding model (DLGE), in which an autoencoder is used to learn the nonlinear embedding of the graph. The specific loss function is defined as follows:
[0256]
[0257] Here V and E represent the set of nodes and edges in the graph, respectively, f(v) represents the low-dimensional embedding of node v, and λ is the regularization coefficient used to adjust the penalty for non-adjacent node pairs, thereby maintaining the basic structure of the graph during dimensionality reduction.
[0258] The lymphedema network is assumed to consist of multiple nodes, each node represents a different lymph node, and the connections between nodes reflect the anatomical or functional connections between lymph nodes. Consider a specific small network:
[0259] Nodes and edges: There are four nodes, and there are edges between each node, indicating the proximity in biological function or anatomy.
[0260] Regularization coefficient λ: Set λ = 0.01 to ensure that the penalty for non-adjacent node pairs is within a reasonable range.
[0261] The feature vector of a node is represented in a high-dimensional space:
[0262] f(v1)=[1.0,2.0], f(v2)=[1.1,1.9], f(v3)=[0.9,2.1], f(v4)=[1.0,2.0]
[0263] For each pair of adjacent nodes (i, j), calculate ||f(v i )-f(v j )|| 2 . For v1 and v2:
[0264] ||f(v1)-f(v2)|| 2 =||[1.0,2.0]-[1.1,1.9]|| 2 =(1.0-1.1) 2 +(2.0-1.9) 2 =0.01+0.01
[0265] =0.02
[0266] Calculate the loss of non-adjacent node pairs, for non-adjacent nodes v1 and v3:
[0267]
[0268] By integrating the losses of all node pairs, we can get the final loss value of the entire network, and adjust the value of f(v) accordingly to minimize the loss, thereby learning the optimal low-dimensional embedding for each node.
[0269] The embodiment further introduces a graph structure preserving loss function (GSPL) to ensure that the topological structure and node relationships of the original graph are maintained during the nonlinear dimensionality reduction process of graph embedding. The formula is as follows:
[0270]
[0271] Among them, ω ij is the importance weight of the node pair (i, j) learned based on the graph attention layer, x i and x j is the original feature vector, g(i) and g(j) are the corresponding low-dimensional representations.
[0272] Suppose you are working on a graph involving lymphedema research, where each node represents a lymph node and has a specific biomarker feature vector. Select a few nodes and connections to specifically show how to implement GSPL.
[0273] Node feature vector (x) and low-dimensional representation (g):
[0274] x1=[1.0,2.0],x2=[1.1,2.1],x3=[0.9,1.9]
[0275] Assuming that the low-dimensional representation has been calculated through the graph embedding process, we get g1 = [0.5, 1.0], g2 = [0.6, 1.1], g3 = [0.4, 0.9]
[0276] Weight setting (ω): Set all weights to 1.
[0277] Compute the difference between the original distance and the low-dimensional representation distance between all pairs of nodes in the graph, and then apply weights:
[0278] For the node pair (1,2):
[0279] Original distance:
[0280] Low-dimensional distance:
[0281] GSPL Contributions:
[0282] Similarly, calculate the other node pairs 1,3 and 2,3, and add up these values to get the total GSPL value.
[0283] All low-dimensional representations correctly preserve the distances between the original feature vectors, and the overall GSPL will be very low, indicating that the embedding is good and maintains the structural integrity of the graph.
[0284] Embodiment 7:
[0285] The embodiment implements a dynamic graph convolutional network (DGCN) to be able to adaptively handle node characteristics that change over time. The network uses a dynamic graph convolution layer to continuously update node representations, thereby effectively detecting abnormal changes in the lymphedema pathology network. The update function is defined as:
[0286]
[0287] Where: t represents the time step; v is the current node; is the neighbor set of node v; W t and b t are the weight matrix and bias vector at time step t respectively; h t-1 (u) is the embedding representation of node u at time t-1; σ is the standard deviation of the Gaussian kernel, which is used to model the interaction strength between nodes.
[0288] The setting is dealing with a dynamic network containing lymph nodes, where each node is connected to other nodes, representing interactions that influence the development of pathological states.
[0289] Parameter settings:
[0290] Assume that there are three main lymph node nodes in the network, each of which is connected to the other two;
[0291] Set σ = 1.0 to balance the effect of distance;
[0292] The weight W at time step t t and bias b t Initialized and learned during training.
[0293] Assume that the initial feature vector of each node is as follows, and the embedding representation at time t-1:
[0294] x1=[1.0,2.0],h t-1 (1) = [0.5, 1.0]
[0295] x2=[1.1,2.1],h t-1 (2) = [0.6, 1.1]
[0296] x3=[0.9,1.9],h t-1 (3) = [0.4, 0.9]
[0297] At time step t, an updated embedding representation is computed for each node:
[0298] For node v1, calculate:
[0299]
[0300] Set ||x2-x1|| 2 = 0.02 and ||x3-x1|| 2 =0.02.
[0301] This computational approach not only reflects the dynamic relationships between nodes, but also updates their features by time step to capture their changes during the pathological process, thus playing a key role in the early detection and analysis of the development of lymphedema.
[0302] The embodiment further applies a graph-based anomaly detection algorithm (GAD) to identify nodes that deviate significantly from common patterns or expected biomarker behaviors, which indicates specific pathological changes or early signs of disease. The calculation formula for anomaly detection is as follows:
[0303]
[0304] in:
[0305] f(v) and f(u) are the embedding vectors of node v and its neighbor u respectively;
[0306] is the neighbor set of node v;
[0307] p is a positive real number used to adjust the sensitivity and discrimination of anomaly detection. The actual p value can be selected from the range of 1 to 3 to adjust according to the characteristics of the actual data and the required sensitivity.
[0308] A simplified lymph node network is set, which contains four nodes, each node represents a different lymph node, and each node has its specific embedding representation, which is obtained by the deep graph embedding method in the previous step.
[0309] Set the embedding vector as follows:
[0310] f(v1)=[1.2,0.8]
[0311] f(v2)=[1.1,0.9]
[0312] f(v3)=[1.3,0.7]
[0313] f(v4)=[0.9,1.0]
[0314] We set p=2 to improve the sensitivity of anomaly detection.
[0315] Take node v1 as an example, and set the neighbor nodes to include v2 and v3:
[0316] Compute the embedding difference between v1 and its neighbors:
[0317] ||f(v1)-f(v2)|| 2 =(1.2-1.1) 2 +(0.8-0.9) 2 =0.01+0.01=0.02
[0318] ||f(v1)-f(v3)|| 2 =(1.2-1.3) 2 +(0.8-0.7) 2 =0.01+0.01=0.02
[0319] Apply the anomaly detection formula:
[0320]
[0321] In this way, a value of d(v1) that is significantly higher than other nodes indicates that v1 exhibits abnormal embedding changes, indicating pathological changes associated with this node.
[0322] The last embodiment combines time series analysis and prediction models to process and predict the pathological progression of lymphedema. The time series graph data is processed by the long short-term memory network (LSTM), effectively capturing and utilizing dynamic biomarkers and pathological states that change over time, thereby providing an accurate analysis tool for disease management and prognosis of lymphedema. The formula is as follows:
[0323] h t =LSTM(h t-1 ,α·h t-1 +(1-α)·x t )
[0324] in:
[0325] h t is the hidden state at time t;
[0326] x t is the input feature at time t;
[0327] α is a learning rate between 0 and 1, which is used to balance the influence of historical information and current input.
[0328] A time series dataset involving the pathological status of lymph nodes is provided, and each time point contains biomarker data from different lymph nodes.
[0329] Parameter settings:
[0330] Assume that each time point t in the data set includes the corresponding lymph node feature vector x t ,For example:
[0331] x1=[0.8,1.2]
[0332] x2=[0.85,1.25]
[0333] x3=[0.9,1.3]
[0334] α is set to 0.5, which means that the model will consider the impact of historical and current inputs equally.
[0335] The initial hidden state h0 is initialized to a small random value based on previous data.
[0336] For each time step t, the hidden state is updated using the LSTM formula:
[0337] h1=LSTM(h0,0.5·h0+0.5·x1)
[0338] h2=LSTM(h1,0.5·h1+0.5·x2)
[0339] h3=LSTM(h2,0.5·h2+0.5·x3)
[0340] In this process, LSTM fine-tunes the information flow by adjusting its response to historical states and new inputs through its internal structure (forget gate, input gate, output gate).
[0341] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.
Claims
1. A method for constructing a lymphedema pathology network based on graph neural network, characterized in that The method includes the following steps: firstly, medical imaging data, clinical records and biomarker data of the patient's lymphatic system are collected, and key features are extracted using image processing technology and structured into nodes and edges of a graph, where nodes represent lymph nodes and edges represent paths or interactions of lymphatic flow; Then, based on the anatomical structure and physiological function of the lymphatic system, the topological structure of the graph is defined, and multimodal data fusion technology is introduced to combine imaging data and biomarker data to enhance the information accuracy of the graph network; Then, a multi-layer graph convolutional network model is designed to learn the patterns of lymph nodes and their connections, and an attention mechanism is introduced to identify and emphasize nodes and edges that play a role in the development of lymphedema. Then, deep learning is performed using graph convolutional networks to extract hidden features of nodes and edges, and graph embedding techniques are applied to transform high-dimensional graph data into low-dimensional space for classification or clustering analysis. Finally, a graph-based anomaly detection algorithm was developed to identify pathological changes that were significantly different from common patterns and combined with time series analysis to predict the pathological progression of lymphedema.
2. The method for constructing a lymphedema pathology network based on a graph neural network according to claim 1, characterized in that The method of extracting key features using image processing technology comprises the following steps: S1. First, adopt a unified data processing framework: Among them, G and slice represent processing functions of different data sources including CT and MRI imaging data, blood biochemical indicators, and patient symptom records, which are used to synchronously process and integrate multiple types of data; S2, then apply the image processing algorithm through the function: where v n is the value of each pixel in the image, is the average pixel value, which is used to identify and quantify the size, shape, and distribution characteristics of lymph nodes; S3, then build the algorithm through the graph: Where E is the set of edges, α ij is the edge weight, k i and k j It is the characteristic value of the nodes connecting the edges, which is used to convert the extracted features into nodes in the graph, i.e., representing lymph nodes, and edges, i.e., representing the physiological relationship or signal transmission path between nodes.
3. The method for constructing a lymphedema pathology network based on a graph neural network according to claim 1, characterized in that Defining the topological structure of the graph comprises the following steps: S1. First, use the graph topology generation algorithm: Where N represents nodes, L represents edges, and L i is a connection specified based on anatomical structure, A i is the corresponding physiological functional connection, d is the distance between nodes, and σ is the normalization parameter used to construct the nodes and edges of the graph according to the anatomical location and physiological function of the lymph nodes; S2, then apply the multimodal deep fusion network through the function: Where x, y, z represent the input data of different modes, w k and v j It is the weight learned from the data, which is used to integrate information from different sources including MRI, CT scans, and biochemical markers to enhance the accuracy of the final graph structure data.
4. The method for constructing a lymphedema pathology network based on a graph neural network according to claim 1, characterized in that The multi-layer graph convolutional network model construction process: First, we introduce and implement the adaptive layer structure of the multi-layer graph convolutional network GCN, using the calculation formula: Where G represents the overall structure of the graph, E represents the edge set in the graph, and x i and x i-1 Represent the feature vectors of consecutive nodes, α i is the inter-layer adaptability factor, ∈ is the smoothing constant, which is used to adjust the structure of the network layer to adapt to the anatomical characteristics and pathological changes of the lymph nodes; Secondly, the graph-based attention mechanism is introduced, and the calculation formula is: where u and v are nodes in the graph, f(u) and f(v) represent the feature vectors of nodes u and v, β is the learned attention concentration parameter, and 6 is a regularization constant used to emphasize nodes and edges that are critically related to the development of lymphedema.
5. The method for constructing a lymphedema pathology network based on a graph neural network according to claim 1, characterized in that The method for extracting hidden features of nodes and edges adopts the steps of: S1. First, we design and implement an adaptive layer structure of a multi-layer graph convolutional network (GCN). We adapt the connectivity and feature diversity of nodes by adjusting the graph convolution operation of the convolution kernel parameters. The convolution kernel size and feature extraction method of each layer are dynamically adjusted according to the local structure of the graph to capture the biomarkers and pathological characteristics between lymph nodes. S2. Further integrate multi-scale feature extraction strategies to capture biomarker features from micro to macro at different levels, and introduce cross-layer connections to enhance information flow and feature reuse to improve the accuracy of classification and clustering; S3. Finally, a nonlinear dimensionality reduction method based on deep learning is applied for graph embedding. Designed for graph data, while maintaining the topological structure of the graph and the relationship between nodes, a high-dimensional to low-dimensional mapping is constructed to facilitate subsequent classification or clustering analysis.
6. The method for constructing a lymphedema pathology network based on a graph neural network according to claim 5, characterized in that The adaptive layer structure adjusts the function through the dynamic convolution kernel: Implementation, where G represents the overall structure of the graph, E represents the edge set, and α vu represents the adaptive weight on the edge (v, u), x v and x u denote the feature vectors of nodes v and u, respectively. θ is a parameter adjusted based on the topology of the graph and the diversity of node features to adapt to the feature extraction requirements of each layer, especially in terms of the biomarkers of lymph nodes and their interactions. Furthermore, the multi-scale feature integration technology is used to effectively capture the specific biomarkers and pathological features of lymphedema, using the formula: Among them l represents the output feature matrix of the lth layer, W k represents the weight matrix from layer k to layer l, and σ is the nonlinear activation function used to integrate information at multiple levels of the graph and extract biomarkers.
7. The method for constructing a lymphedema pathology network based on a graph neural network according to claim 5, characterized in that The integrated multi-scale feature extraction strategy includes the following steps: S1. First, implement the multi-scale graph convolutional network M-GCN through the function: Where v represents the current node, N k (v) is the set of neighbor nodes of node v at the kth scale, ω k is the learning weight of the kth scale, x u is the feature of the neighbor nodes, σ is the activation function designed to capture the biomarker features from micro to macro at different levels; S2, then introduce cross-layer connections to enhance information flow and feature reuse, specifically through the formula: Implementation, where F l is the output feature of the lth layer, H i is the output feature of the i-th layer, γ li It is the cross-layer connection weight from the i-th layer to the l-th layer, which is used to improve the accuracy of classification and clustering.
8. The method for constructing a lymphedema pathology network based on a graph neural network according to claim 5, characterized in that The nonlinear dimensionality reduction method for constructing a graph embedding framework includes: S1. First, we implement the deep learning driven graph embedding model DLGE, which learns the nonlinear embedding of the graph through the autoencoder, specifically by minimizing the function: Where V and E represent the node set and edge set in the graph respectively, f(v) represents the low-dimensional embedding of node v, and λ is the regularization coefficient, which is used to adjust the penalty strength of non-adjacent node pairs to maintain the topological structure of the graph and the relationship between nodes in the process of high-dimensional to low-dimensional mapping; S2. Then introduce the graph structure preservation loss function GSPL, by calculating: where ω ij is the importance weight of the node pair (i, j) learned based on the graph attention layer, x i and x j is the original feature vector, g(i) and g(j) are the corresponding low-dimensional representations.
9. The method for constructing a lymphedema pathology network based on a graph neural network according to claim 1, characterized in that The graph-based anomaly detection algorithm consists of: S1. First, implement the dynamic graph convolution network DGCN, using the dynamic graph convolution layer, through the function: To update the node representation, where t represents the time step, v is the current node, is the neighbor set of node v, W t and b t are the weight matrix and bias vector at time step t, respectively, h t-1 (u) is the embedding representation of node u at time t-1, σ is the standard deviation of the Gaussian kernel, which is used to model the interaction strength between nodes; S2, then apply the graph-based anomaly detection algorithm GAD, by calculating: to identify anomalies, where f(v) and f(u) are the embedding vectors of node v and its neighbor u, respectively, and p is a positive real number selected to increase the sensitivity and discrimination of anomaly detection; S3. Finally, combining time series analysis and prediction model, the time series graph data is processed through the long short-term memory network LSTM. The specific implementation formula is: h t =LSTM(h t-1 ,α·h t-1 +(1-α)·x t ) where h t is the hidden state at time t, x t is the input feature, and α is a learning rate between 0 and 1, which is used to balance the influence of historical information and current input to predict the pathological progression of lymphedema.
Citation Information
Patent Citations
Extensible multi-level graph neural network model based on multi-modal image data
CN115393269A
Network embedding method and system based on graph neural network
CN117493617A
Breast cancer clinical data analysis diagnosis and treatment platform based on artificial intelligence large language model
CN118016280A
Method and system for evaluating condition of sepsis patient based on multi-modal data fusion
CN118280579A
Digitized monitoring and early warning method and system for plant diseases and insect pests
CN118334582A
Cited By
Personalized hypertensive nephropathy risk assessment method and system based on graph convolutional network
CN119673431A
Gastric cancer postoperative prognosis evaluation system based on nonlinear Cox
CN122050843A
A nonlinear Cox-based postoperative prognostic assessment system for gastric cancer
CN122050843B
